[Ltrace-devel] DWARF test suite

[Dima Kogan]

Hi. Apologies for not replying earlier. I've been busy. I rebased my
tree on top of the latest HEAD, and added a new dwarf test fix.



Petr Machata <pmachata at redhat.com> writes:

> bae1df98b (aliased symbol was breaking) looks OK, though that if (X)
> return CBS_STOP; else return CBS_CONT; can be rewritten more concisely
> as return CBS_STOP_IF (X).  Let me know if you want to update it.

I included this change into the appropriate patch
(edb44b33fc431ac59bdccdcd56ae038e1b1adf99).


> I appreciate the tests, this should help us keep the Dwarf part stable.
> However I would like to see a C-only elementary test, possibly with
> optional C++ part.  C++ is not a hard dependency of ltrace.  It is OK if
> those parts use C++11, I understand that you need to test those parts of
> Dwarf output.  I'm however opposed to just enabling C++11 en bloc for
> all C++ files.  Personally, I'd prefer that you just change this:
>
>         [ltrace_compile_shlib $libsrc $lib_so debug ]
>
> ... to this:
>
>         [ltrace_compile_shlib $libsrc $lib_so
>                 [list debug c++ additional_flags=-std=gnu++11]]
>
> Though if you want to put the logic that just enables c++ in there like
> you do in d983fbdf, that makes sense to me as well.  In that case,
> please update demangle.exp so that it doesn't list the c++ option
> unnecessarily.  But in any case, please keep the C++11 flags private to
> your test case.

OK. What you suggest doesn't really work because then everything in my
tests will be built as c++, instead of just the .cc. I can make us both
happy, but let me understand specifically what the complaint is. Are you
opposed to the full patch to ltrace.exp (build all .cc and .cpp as c++
source), or just the fact that I'm building them as a gnu++11 flavor of
c++? If I removed the gnu++11 part from the ltrace.exp patch, and put
that into dwarf.exp somehow, then we're good?

dima